This is another optimization to help improve the baseline overhead
of async ops. It shaves off ~55ns/op or ~7% of the current total
async op overhead.
It achieves these gains by taking advantage of the sequential
nature of promise IDs and optimistically stores them sequentially
in a pre-allocated circular buffer and fallbacks to the promise Map
for slow to resolve promises.
- Improves op performance.
- Handle op-metadata (errors, promise IDs) explicitly in the op-layer vs
per op-encoding (aka: out-of-payload).
- Remove shared queue & custom "asyncHandlers", all async values are
returned in batches via js_recv_cb.
- The op-layer should be thought of as simple function calls with little
indirection or translation besides the conceptually straightforward
serde_v8 bijections.
- Preserve concepts of json/bin/min as semantic groups of their
inputs/outputs instead of their op-encoding strategy, preserving these
groups will also facilitate partial transitions over to v8 Fast API for the
"min" and "bin" groups
This commit moves implementation of bin ops to "deno_core" crates
as well as unifying logic between bin ops and json ops to reuse
as much code as possible (both in Rust and JavaScript).
This commit rewrites "dispatch_minimal" into "dispatch_buffer".
It's part of an effort to unify JS interface for ops for both json
and minimal (buffer) ops.
Before this commit "minimal ops" could be either sync or async
depending on the return type from the op, but this commit changes
it to have separate signatures for sync and async ops (just like
in case of json ops).
In case JavaScript throws an unhandled error, part of the "shared_queue" could
be still unprocessed.
If this is the case; throw the js runtime error instead of asserting on the
queue size not being 0.
This commit rewrites implementation of "JsRuntime::mod_evaluate".
Event loop is no longer polled automatically and users must manually
drive event loop forward after calling "mod_evaluate".
Co-authored-by: Nayeem Rahman <nayeemrmn99@gmail.com>
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This commit rewrites "JsRuntime::poll" function to fix a corner case that
might caused "overflown_response" to be overwritten by other overflown response.
The logic has been changed to allow returning multiple overflown response
alongside responses from shared queue.
This commit adds two new "Deno.core" APIs:
- "Deno.core.serialize"
- "Deno.core.deserialize"
These APIs are used to provide structured cloning of values
and will be used for further web worker implementation.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
This PR makes json_op_sync/async generic to all Deserialize/Serialize types
instead of the loosely-typed serde_json::Value. Since serde_json::Value
implements Deserialize/Serialize, very little existing code needs to be updated,
however as json_op_sync/async are now generic, type inference is broken in some
cases (see cli/build.rs:146). I've found this reduces a good bit of boilerplate,
as seen in the updated deno_core examples.
This change may also reduce serialization and deserialization overhead as serde
has a better idea of what types it is working with. I am currently working on
benchmarks to confirm this and I will update this PR with my findings.
This commit rewrites initialisation of the "shared queue" and
in effect prevents from double execution of "core/core.js" and
"core/error.js".
Previously both of these files were executed every time a "JsRuntime"
was created. That lead to a situation where one copy of each script
was included in the snapshot and then another copy would be
executed after loading the snapshot.
Effectively "JsRuntime::shared_init" was removed; instead execution
of those scripts and actual initialisation of shared queue
was split into two helper functions: "JsRuntime::js_init" and
"JsRuntime::share_queue_init".
Additionally stale TODO comments were removed.
Fixes the following runtime error for me when benchmarking:
thread 'main' panicked at 'called `Result::unwrap()` on an `Err`
value: Error: Unregistered error class: "Error"
Connection reset by peer (os error 104)
Classes of errors returned from ops should be registered via
Deno.core.registerErrorClass().
at processResponse (deno:core/core.js:219:13)
at Object.jsonOpAsync (deno:core/core.js:240:12)
at async read (http_bench_json_ops.js:29:21)
at async serve (http_bench_json_ops.js:45:19)',
core/examples/http_bench_json_ops.rs:260:28
Implementors of `deno_core::JsRuntime` might want to do additional actions
during each turn of event loop, eg. `deno_runtime::Worker` polls inspector,
`deno_runtime::WebWorker` receives/dispatches messages from/to worker host.
Previously `JsRuntime::mod_evaluate` was implemented in such fashion that it
only polled `JsRuntime`'s event loop. This behavior turned out to be wrong
in the example of `WebWorker` which couldn't receive/dispatch messages because
its implementation of event loop was never called.
This commit rewrites "mod_evaluate" to return a handle to receiver that resolves
when module's promise resolves. It is now implementors responsibility to poll
event loop after calling `mod_evaluate`.
This commit migrates all ops to use new resource table
and "AsyncRefCell".
Old implementation of resource table was completely
removed and all code referencing it was updated to use
new system.
This commit moves Deno JS runtime, ops, permissions and
inspector implementation to new "deno_runtime" crate located
in "runtime/" directory.
Details in "runtime/README.md".
Co-authored-by: Ryan Dahl <ry@tinyclouds.org>
This commit does major refactor of "Worker" and "WebWorker",
in order to decouple them from "ProgramState" and "Flags".
The main points of interest are "create_main_worker()" and
"create_web_worker_callback()" functions which are responsible
for creating "Worker" and "WebWorker" in CLI context.
As a result it is now possible to factor out common "runtime"
functionality into a separate crate.
This commit adds "Deno.core.createPrepareStackTrace". This function
was moved from "cli/rt/40_error_stack.js" to unify handling of stack frames in core
(before this PR there was implicit dependency on logic in "core/error.rs::JsError").
Unfortunately formatting logic must still be duplicated in "cli/error.js::PrettyJsError"
to provide coloring, but currently there's no solution to this problem.
"createPrepareStackTrace" can accept a single argument; a function that takes
a location and provides source mapped location back.
This commit disables source mapping of errors
for standalone binaries. Since applying source
maps relies on using file fetcher infrastructure
it's not feasible to use it for standalone binaries
that are not supposed to use that infrastructure.
This commit adds `FsModuleLoader` to `deno_core`, which implements
`ModuleLoader` trait. It is used when creating a runtime that supports
module loading from filesystem.
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
Fixes panic occurring in worker when "self.close()" is called
at the top level, ie. worker shuts down while
module evaluation promise hasn't yet resolved.
This commit refactors "deno_core::Modules" structure to not depend on
"get_identity_hash" function to identify modules, but instead use default
hash implementation.
Make JSRuntime::new() accept a custom v8::CreateParams object to tune
the v8::Isolate it creates.
Subsumes the functionality of HeapLimits, which I therefore removed.
This commit changes implementation of top-level-await in "deno_core".
Previously promise returned from module evaluation was not awaited,
leading to out-of-order execution of modules that have TLA. It's been
fixed by changing "JsRuntime::mod_evaluate" to be an async function
that resolves when the promise returned from module evaluation also
resolves. When waiting for promise resolution event loop is polled
repeatedly, until there are no more dynamic imports or pending
ops.
* Revert "refactor: Worker is not a Future (#7895)"
This reverts commit f4357f0ff9.
* Revert "refactor(core): JsRuntime is not a Future (#7855)"
This reverts commit d8879feb8c.
* Revert "fix(core): module execution with top level await (#7672)"
This reverts commit c7c7677825.
This commit rewrites deno_core::JsRuntime to not implement Future
trait.
Instead there are two separate methods:
- JsRuntime::poll_event_loop() - does single tick of event loop
- JsRuntime::run_event_loop() - runs event loop to completion
This commit renames occurrences of "isolate" variable name
to "js_runtime". This was outstanding debt after renaming
deno_core::CoreIsolate to JsRuntime.
This commit fixes implementation of top level await in "deno_core".
Previously promise returned from module execution was ignored causing to execute
modules out-of-order.
With this commit promise returned from module execution is stored on "JsRuntime"
and event loop is polled until the promise resolves.
This makes use of a default referrer when its empty in repl mode so that
dynamic imports work in the global evaluation context.
Co-authored-by: Bartek Iwanczuk <biwanczuk@gmail.com>
This commit adds support for stack traces in "deno_core".
Implementation of "Display" trait for "JsError" has been updated
and in consequence "deno_core::js_check" became obsolete and
removed.
- remove "CliState.workers" and "CliState.next_worker_id", instead
store them on "OpState" using type aliases.
- remove "CliState.global_timer" and "CliState.start_time", instead
store them on "OpState" using type aliases.
- remove "CliState.is_internal", instead pass it to Worker::new
- move "CliState::permissions" to "OpState"
- move "CliState::main_module" to "OpState"
- move "CliState::global_state" to "OpState"
- move "CliState::check_unstable()" to "GlobalState"
- change "cli_state()" to "global_state()"
- change "deno_core::ModuleLoader" trait to pass "OpState" to callbacks
- rename "CliState" to "CliModuleLoader"
Moves op_close and op_resources to deno_core::ops and exports them.
Adds serde dependency to deno_core and reexports it.
Moves JS implementation of those ops to Deno.core and reexports them in Deno.
Removes:
- "deno_core::StartupData"
- "deno_core::Script"
- "deno_core::OwnedScript"
Changes to "JsRuntime":
- remove "new_with_loader()"
- remove "with_heap_limits()"
- rename "IsolateOptions" to "RuntimeOptions" and make public
- "JsRuntime::new()" takes "RuntimeOptions" as a single param
Provides a concrete state type that can be dynamically added. This is necessary for op crates.
* renames BasicState to OpState
* async ops take `Rc<RefCell<OpState>>`
* sync ops take `&mut OpState`
* removes `OpRegistry`, `OpRouter` traits
* `get_error_class_fn` moved to OpState
* ResourceTable moved to OpState
This commit changes "deno info" subcommand logic.
- Modules are no longer loaded into V8 isolate - analysis
is done using ModuleGraph.
- Removed deno_core::Deps structure.
- Modules are no longer type-checked and transpiled -
"compiled" file is shown only if it is already available.
- Added number of unique dependencies for root module.
- Changed tree output:
- file size is shown next to the dependency
- repeated dependencies are marked with "*"
- used less spaces in prefix to save terminal width
deno_core/
- rename core_isolate.rs to runtime.rs
- rename CoreIsolate to JsRuntime
- rename JSError to JsError
- rename JSStackFrame to JsStackFrame
cli/
- update references from deno_core::CoreIsolate to deno_core::JsRuntime
- rename deno_core::JSError to deno_core::JsError
- rename fmt_errors::JSError to fmt_errors::JsError
Removes unnecessary wrappers around ResourceTable and CoreIsolate.
Importantly: now uses the resource table inside CoreIsolateState rather than a different one.
Removes derive_deref dependency
The following crates were _not_ upgraded to avoid having multiple
versions of the same crate in the dependency tree:
* tokio-tungstenite v0.10.1 -> v0.11.0
* swc_common v0. 8.0 -> v0. 9.1
* swc_ecmascript v0. 1.0 -> v0. 3.0
* webpki-roots v0.19.0 -> v0.20.0
* nix v0.17.0 -> v0.18.0
This commit adds alternate dispatch method to core JS API.
"Deno.core.dispatchByName()" works like "Deno.core.dispatch()",
but takes op name instead of op id as a first argument.
This commit fixes problems with source maps in Chrome Devtools
by substituting source map URL generated by TS compiler with
actual file URL pointing to DENO_DIR.
Dummy value of "source_map_url" has been removed from
"ScriptOrigin".
Also fixes lock file which used compiled source code to generate
lock hash; it now uses source code of the file that is
being compiled.
This PR adds prepare_load hook method to ModuleLoader trait. It allows implementors to perform preparation work before starting actual module loading into isolate. It's meant to be used in CLI; where "transpilation" step will be explicitly performed during prepare_load instead of doing it adhoc for each module if needed.
Co-authored-by: Filipe Schenkel de Souza <filipe.schenkel@azion.com>
Co-authored-by: Douglas Caetano dos Santos <douglas.santos@azion.com>
Co-authored-by: João Avelino Bellomo Filho <joao.avelino@azion.com>
This is a first pass implementation which is still missing several important
features:
- support for --inspect-brk (#4503)
- support for source maps (#4501)
- support for piping console.log to devtools console (#4502)
Co-authored-by: Bert Belder <bertbelder@gmail.com>
Co-authored-by: Matt Harrison <mt.harrison86@gmail.com>
Co-authored-by: Bartek Iwańczuk <biwanczuk@gmail.com>
* Remove DENO_BUILD_MODE and DENO_BUILD_PATH
Also remove outdated docs related to ninja/gn.
* fix
* remove parameter to build_mode()
* remove arg parsing from benchmark.py
* add tests for "Deno.core.encode" and "Deno.core.decode" for empty inputs
* use "Deno.core.encode" in "TextEncoder"
* use "Deno.core.decode" in "TextDecoder"
* remove "core_decode" and "core_encode" benchmarks
This commits add two new methods to "Deno.core" namespace: "encode" and "decode".
Those methods are bound in Rust to provide a) fast b) generally available of encoding and decoding UTF-8 strings.
Both methods are now used in "cli/js/dispatch_json.ts".
The V8 documentation explicitly states that SnapshotCreator::CreateBlob()
should not be called from within a HandleScope.
Additionally, this patch removes some non-functional error handling code
from the deno_core::Isolate::snapshot() method.
* establish basic event loop for workers
* make "self.close()" inside worker
* remove "runWorkerMessageLoop() - instead manually call global function
in Rust when message arrives. This is done in preparation for structured clone
* refactor "WorkerChannel" and use distinct structs for internal
and external channels; "WorkerChannelsInternal" and "WorkerHandle"
* move "State.worker_channels_internal" to "Worker.internal_channels"
* add "WorkerEvent" enum for child->host communication;
currently "Message(Buf)" and "Error(ErrBox)" variants are supported
* add tests for nested workers
* add tests for worker throwing error on startup
This change simplifies how we execute V8. Previously V8 Isolates jumped
around threads every time they were woken up. This was overly complex and
potentially hurting performance in a myriad ways. Now isolates run on
their own dedicated thread and never move.
- blocking_json spawns a thread and does not use a thread pool
- op_host_poll_worker and op_host_resume_worker are non-operational
- removes Worker::get_message and Worker::post_message
- ThreadSafeState::workers table contains WorkerChannel entries instead
of actual Worker instances.
- MainWorker and CompilerWorker are no longer Futures.
- The multi-threaded version of deno_core_http_bench was removed.
- AyncOps no longer need to be Send + Sync
This PR is very large and several tests were disabled to speed
integration:
- installer_test_local_module_run
- installer_test_remote_module_run
- _015_duplicate_parallel_import
- _026_workers
* move is_dyn_import argument from Loader::resolve to Loader::load - it was always kind of strange that resolve() checks permissions.
* change argument type from &str to &ModuleSpecifier where applicable
Ref #3712. This change allowed the deno_typescript crate to reference
cli/js/lib.deno_runtime.d.ts which breaks "cargo package". We intend to
reintroduce a revised version of this patch later once "cargo
package" is working and tested.
This reverts commit 737ab94ea1.
* split ops/worker.rs into ops/worker_host.rs and ops/web_worker.rs
* refactor js/workers.ts and factor out js/worker_main.ts - entry point for WebWorker runtime
* BREAKING CHANGE: remove support for blob: URL in Worker
* BREAKING CHANGE: remove Deno namespace support and noDenoNamespace option in Worker constructor
* introduce WebWorker struct which is a stripped down version of cli::Worker
* remove Isolate.current_send_cb_info
* remove DenoBuf
* remove Isolate.shared_ab
* port Isolate.shared_response_buf (last bit not ported from libdeno)
* add some docs for Isolate and EsIsolate
* refactored RecursiveLoad - it was renamed to RecursiveModuleLoad, it does not take ownership of isolate anymore - a struct implementing Stream that yields SourceCodeInfo
* untangled module loading logic between RecursiveLoad and isolate - that logic is encapsulated in EsIsolate and RecursiveModuleLoad, where isolate just consumes modules as they become available - does not require to pass Arc<Mutex<Isolate>> around anymore
* removed EsIsolate.mods_ in favor of Modules and moved them inside EsIsolate
* EsIsolate now requires "loader" argument during construction - struct that implements Loader trait
* rewrite first methods on isolate as async
Some tests were silently failing after #3358 and #3434 because pool.spawn_ok
was used which doesn't panic on errors. For reference, the failure looked like this:
thread '<unnamed>' panicked at 'assertion failed: match isolate.poll_unpin(cx) { Poll::Ready(Ok(_)) => true, _ => false, }', core/isolate.rs:1408:7
After landing #3358 the benchmarks exploded indicating problems with workers and deno_core_http_bench.
This PR dramatically fixes thread/syscall count that showed up on benchmarks. Thread count is not back to previous levels but difference went from hundreds/thousands to about ~50.
- removes global `RESOURCE_TABLE` - resource tables are now created per `Worker`
in `State`
- renames `CliResource` to `StreamResource` and moves all logic related
to it to `cli/ops/io.rs`
- removes `cli/resources.rs`
- adds `state` argument to `op_read` and `op_write` and consequently adds
`stateful_minimal_op` to `State`
- IMPORTANT NOTE: workers don't have access to process stdio - this is
caused by fact that dropping worker would close stdout for process
(because it's constructed from raw handle, which closes underlying file
descriptor on drop)
This patch does not work with the recent bundler changes (#3325).
Unfortunately I didn't merge master before landing this patch. It has
something to do with console.log not working inside the compiler worker.
This reverts commit fd62379eaf.
- removes global `RESOURCE_TABLE` - resource tables are now created per `Worker`
in `State`
- renames `CliResource` to `StreamResource` and moves all logic related
to it to `cli/ops/io.rs`
- removes `cli/resources.rs`
- adds `state` argument to `op_read` and `op_write` and consequently adds
`stateful_minimal_op` to `State`
- IMPORTANT NOTE: workers don't have access to process stdio - this is
caused by fact that dropping worker would close stdout for process
(because it's constructed from raw handle, which closes underlying file
descriptor on drop)
Towards simplifying (or better removing entirely) the CoreResource
trait. Resources should be any bit of privileged heap allocated memory
that needs to be referenced from JS, not very specific trait
implementations. Therefore CoreResource should be pushed towards being
as general as possible.
- Fixes cargo publish on deno_typescript, deno_cli_snapshots, and
deno_cli.
- Combines cli_snapshots and js into one directory.
- Extracts TS version at compile time rather than runtime
- Bumps version awkwardly - it was necessary to test end-to-end
publishing. Sorry.
- Adds git submodule deno_typescript/typescript
Instead of using core/snapshot_creator.rs, instead two crates are
introduced which allow building the snapshot during build.rs.
Rollup is removed and replaced with our own bundler. This removes
the Node build dependency. Modules in //js now use Deno-style imports
with file extensions, rather than Node style extensionless imports.
This improves incremental build time when changes are made to //js files
by about 40 seconds.
Deno.core.dispatch() used to push the "control" buf onto the shared
array buffer before calling into V8, with the idea that it was one less
argument to parse. Turns out there is no more overhead passing the
control ArrayBuffer directly over. Furthermore this optimization was
making the refactors outlined in #2730 more complex. Therefore it is
being removed.
The rules are now as follows:
* In `import` statements, as mandated by the WHATWG specification,
the import specifier is always treated as a URL.
If it is a relative URL, it must start with either / or ./ or ../
* A script name passed to deno as a command line argument may be either
an absolute URL or a local path.
- If the name starts with a valid URI scheme followed by a colon, e.g.
'http:', 'https:', 'file:', 'foo+bar:', it always interpreted as a
URL (even if Deno doesn't support the indicated protocol).
- Otherwise, the script name is interpreted as a local path. The local
path may be relative, and operating system semantics determine how
it is resolved. Prefixing a relative path with ./ is not required.
1. Separate Snapshot and Script StartupData functions based on cfg "no-snapshot-init"
2. Replace deprecated Once::ONCE_INIT with Once::new (https://github.com/rust-lang/rust/pull/61757)
3. Elide lifetime
4. Fix typos
It's unnecessary indirection and is preventing the ability to easily
pass isolate references into the dispatch and dyn_import closures.
Note: this changes how StartupData::Script is executed. It's no longer done
during Isolate::new() but rather lazily on first poll or execution.
This patch makes it so that RecursiveLoad doesn't own the Isolate, so
Worker::execute_mod_async does not consume itself.
Previously Worker implemented Loader, but now ThreadSafeState does.
This is necessary preparation work for dynamic import (#1789) and import
maps (#1921)
* Compiler no longer has its own Tokio runtime. Compiler handles one
message and then exits.
* Uses the simpler ts.CompilerHost interface instead of
ts.LanguageServiceHost.
* avoids recompiling the same module by introducing a hacky but simple
`hashset<string>` that stores the module names that have been already
compiled.
* Removes the CompilerConfig op.
* Removes a lot of the mocking stuff in compiler.ts like `this._ts`. It
is not useful as we don't even have tests.
* Turns off checkJs because it causes fmt_test to die with OOM.
This patch provides a work-around for an apparent V8 bug where
initializing multiple isolates concurrently leads to a crash on
Windows.
At the time of writing the cause of this crash is not exactly
understood, but it seems to be related to the V8 internal
function win64_unwindinfo::RegisterNonABICompliantCodeRange(),
which didn't exist in older versions of V8.
* In order to prevent ArrayBuffers from getting garbage collected by V8,
we used to store a v8::Persistent<ArrayBuffer> in a map. This patch
introduces a custom ArrayBuffer allocator which doesn't use Persistent
handles, but instead stores a pointer to the actual ArrayBuffer data
alongside with a reference count. Since creating Persistent handles
has quite a bit of overhead, this change significantly increases
performance. Various HTTP server benchmarks report about 5-10% more
requests per second than before.
* Previously the Persistent handle that prevented garbage collection had
to be released manually, and this wasn't always done, which was
causing memory leaks. This has been resolved by introducing a new
`PinnedBuf` type in both Rust and C++ that automatically re-enables
garbage collection when it goes out of scope.
* Zero-copy buffers are now correctly wrapped in an Option if there is a
possibility that they're not present. This clears up a correctness
issue where we were creating zero-length slices from a null pointer,
which is against the rules.
Op dispatch is now dynamically dispatched, so slightly less efficient.
The immeasurable perf hit is a reasonable trade for the API simplicity
that is gained here.
Additionally, instead of polling ops in a loop until none of them are
ready, the isolate will now yield to the task system after delivering
the first batch of completed ops to the javascript side.
Although this makes performance a bit worse (about 15% fewer
requests/second on the 'deno_core_http_bench' benchmark), we feel that
the advantages are worth it:
* It resolves the extremely high worst-case latency that we were seeing
on deno_core_http_bench, in particular when using the multi-threaded
Tokio runtime, which would sometimes exceed a full second.
* Before this patch, the implementation of Isolate::poll() had to loop
through all sub-futures and poll each one of them, which doesn't scale
well as the number of futures managed by the isolate goes up. This
could lead to poor performance when e.g. a server is servicing
thousands of connected clients.
This change is made in preparation for using FuturesUnordered to track
futures that are spawned by the isolate. FuturesUnordered sets up
notififications for every future that it finds to be not ready when
polled, which causes a crash if attempted outside of a task context.
* Moves how snapshots are supplied to the Isolate. Previously they were
given by Behavior::startup_data() but it was only called once at
startup. It makes more sense (and simplifies Behavior) to pass it to the
constructor of Isolate.
* Adds new libdeno type deno_snapshot instead of overloading
deno_buf.
* Adds new libdeno method to delete snapshot deno_snapshot_delete().
* Renames deno_get_snapshot() to deno_snapshot_new().
* Makes StartupData hold references to snapshots. This was implicit when
it previously held a deno_buf but is made explicit now. Note that
include_bytes!() returns a &'static [u8] and we want to avoid
copying that.
Fixes some sed errors introduced in c43cfe.
Unfortunately moving libdeno required splitting build.rs into two parts,
one for cli and one for core.
I've also removed the arm64 build - it's complicating things at this
re-org and we're not even testing it. I need to swing back to it and get
tools/test.py running for it.
This disables a few tests which are broken still:
- tests/error_004_missing_module.test
- tests/error_005_missing_dynamic_import.test
- tests/error_006_import_ext_failure.test
- repl_test test_set_timeout
- repl_test test_async_op
- repl_test test_set_timeout_interlaced
- all of permission_prompt_test
We always pass around Box<[u8]>, and adding this generic is an
unnecessary complication.
Add deno_core_http_bench_test to test.py
sharedQueue works on deno_core_http_bench
